Accurate Dual-Channel Broadband RF Attenuation Measurement System with High Attenuation Capability Using an Optical Fiber Assembly for Optimal Channel Isolation

Abstract

In this study, an accurate attenuation measurement system with high attenuation capability (≥100 dB) is presented, covering a broad radio frequency range from 1 GHz to 25 GHz. The system employs a dual-channel intermediate frequency (IF) substitution method, utilizing a programmable inductive voltage divider (IVD) that provides precise voltage ratios at a 1 kHz operating IF, serving as the primary attenuation standard. To ensure optimal inter-channel isolation, essential for accurate high-attenuation measurements, an optical fiber assembly, consisting of a laser diode, a wideband external electro-optic modulator, and a photodetector, is integrated between the channels. A comprehensive performance evaluation is presented, with particular emphasis on the programmable IVD calibration technique, which achieves an accuracy better than 0.001 dB across all attenuation levels, and on the role of the optical fiber assembly in enhancing isolation, demonstrating levels exceeding 120 dB across the entire frequency range. The system demonstrates measurement capabilities with expanded uncertainties (k = 2) of 0.004 dB, 0.008 dB, and 0.010 dB at attenuation levels of 20 dB, 60 dB, and 100 dB, respectively.

1. Introduction

Attenuation is a fundamental quantity standard in the calibration of radio frequency (RF) measurement instruments, including power meters, spectrum analyzers, and vector network analyzers. These instruments are essential for characterizing circuits, devices, and antennas across the RF spectrum, which spans from 3 kHz to 300 GHz.

A variety of attenuation measurement techniques have been developed and employed in various attenuator calibration systems, including power ratio methods, voltage ratio methods, RF substitution methods, and IF substitution methods [1,2,3]. However, as RF devices become increasingly complex, such as with the use of massive multi-input and multi-output (MIMO) antennas in 5G/6G wireless technology, the demand for attenuation measurement standards that are accurate, traceable, broad-band, and capable of covering a high attenuation range has increased. Consequently, research on advanced attenuation measurement techniques and standards remains an active and essential area [4,5,6,7,8,9,10,11,12,13,14].

Attenuation measurement systems based on the dual channel IF substitution method with null detection provide a promising solution to meet these requirements. This approach offers significant advantages across a wide frequency range by converting all RF attenuation to a single IF and comparing it to a primary attenuation standard at that IF. Moreover, because the null detection technique is less susceptible to noise, it ensures high measurement accuracy even at high attenuation levels. However, in a dual channel measurement system, particularly when measuring high attenuation, channel isolation becomes a critical factor, as it can significantly affect measurement accuracy. A common solution involves cascading multiple isolators or amplifiers, but this can restrict system bandwidth and may not fully eliminate ground loop issues.

In our previous work, we developed a highly accurate attenuation measurement system covering two frequency ranges: 10 MHz to 100 MHz and 100 MHz to 1 GHz [5,15]. These systems are based on the dual-channel IF substitution method with null detection, employing a calibrated inductive voltage divider (IVD) operating at 1 kHz as the reference standard. High channel isolation was achieved by incorporating an optical fiber assembly comprising a light-emitting diode (LED) or laser diode (LD) as the light source, optical fibers, and a photodetector (PD). The RF electrical signal is transmitted through the optical fiber assembly by directly modulating the light sources. These systems have been adopted as the national attenuation standard in Japan and have demonstrated excellent accuracy and stability over more than 15 years of continuous operation.

This paper presents a newly developed precision dual-channel attenuation measurement system that operates over a frequency range from 1 GHz to 25 GHz. The system is designed for the evaluation of both variable and fixed attenuator devices under test (DUTs), supporting a measurement range exceeding 100 dB. To accommodate this frequency range, appropriate instruments and components, such as signal sources, frequency mixers, and other RF devices were carefully selected. The optical fiber assembly redesigned to incorporate an external optical modulator (EOM), enabling optical signal modulation across the full frequency range. High-accuracy measurement of fixed attenuators is achieved by employing a saturation amplifier at the local signal source port of the mixer, effectively suppressing amplitude variations induced by flexible cable movement during measurement. To enhance automation, a programmable IVD is employed, and a precise null-detection technique based on interpolation is implemented. A double-step measurement technique is applied to address high-attenuation scenarios. The performance of the developed system, including key sources of uncertainty, is thoroughly evaluated, and the overall measurement uncertainties are presented.

2. Measurement System

Figure 1 shows a block diagram of the proposed RF attenuation measurement system using the dual channel IF substitution method, operating in the frequency range of 1 GHz to 25 GHz. The system consists of an RF section (highlighted in light blue) and an IF section (in light green), or a 1 kHz null-balance receiver, which employs a seven-decade programable IVD (Model 6430A) as a reference standard. The arrows in solid lines indicate the direction of signal flow. In Channel 1, the RF signal from the RF signal source (Model MG3697C) passes through the DUT and is mixed with the local (LO) signal from the LO signal source (also generated by a Model MG3697C) at the main mixer (Model M40126LK) to produce a 1 kHz IF signal, where the amplitude is proportional to the input RF signal. After sufficient signal amplification, the IF signal passes through the IVD and enters the CH1 port of the lock-in amplifier (LiA; Model 5210), which is used as a precision null detector. In Channel 2, the RF signal is converted into a 1 kHz IF signal at the reference mixer, which is also a Model M40126LK. IF reference signal passes through the IF amplifier and IF phase shifter (PS; Model 5920A) and is applied to the CH2 input port of the LiA for measurement. The signals at the CH1 and CH2 ports of the LiA are kept balanced by adjusting the IVD and PS before and after setting the DUT. Letting the IVD values at the initial and final settings be D_i and D_f, respectively, the attenuation value A is determined in dB by the following:

$$A=-20{log}_{10}{\displaystyle \frac{D_i}{D_f}}$$

(1)

For attenuation measurements up to 60 dB, a single-step technique is employed, leveraging the mixer’s excellent linearity over the −20 dBm to −80 dBm range. For measurements exceeding 60 dB, a double-step technique is adopted, incorporating a gauge block attenuator (GBA) to preserve measurement accuracy by reducing the influence of noise.

In high attenuation measurements, if the isolation between the channels is insufficient, as shown in Figure 1 by the dotted line, unwanted signals flow and cause errors, i.e., internal leakage errors in the measurement results. If we let the signals arriving at the main mixer through the DUT and the internal leakage paths have amplitude of S and L, respectively, then the maximum uncertainty ΔA_L due to the internal leakage in the attenuation measurements can be expressed in dB by:

$$∆A_L=-20{\mathit{log}}_{10}{\left(1\pm {\displaystyle \frac{L}{S}}\right)}^{-1}$$

(2)

To maintain a measurement uncertainty below 0.001 dB, the signal ratio between L and S must be less than 1 × 10⁻⁴, corresponding to an isolation greater than 80 dB. For example, measuring an attenuation of 100 dB requires an isolation level of 180 dB—a value that is extremely difficult to achieve using conventional RF devices. To address this challenge, an optical fiber assembly (highlighted in orange) is implemented between the local source and the reference mixer, serving as an ideal isolator that permits only forward transmission of the RF signal from the local source. In this configuration, the RF signal modulates an optical carrier from a laser diode (LD) via an electro-optic modulator (EOM). The modulated optical signal is transmitted through a single-mode optical fiber to a photodetector (PD), where it is converted back into an RF signal and delivered to the reference mixer. The optical assembly consists of a stabilized Fabry–Pérot LD with a 1550 nm wavelength and 0 dBm output power (Model MU951501A9, Anritsu, Tokyo, Japan), a lithium niobate EOM with a 40 GHz bandwidth (Model T-PIDH1.5-ADC, Sumitomo Osaka Cement, Tokyo, Japan), and a PD with a 25 GHz bandwidth and a 900–1700 nm wavelength range (Model 1414, Newport, Irvine, CA, USA). All components are interconnected using single-mode optical fibers terminated with FC connectors.

High-accuracy fixed attenuation measurements are achieved by employing a saturation amplifier (SAT AMP, Model A0126EZP, Marki Microwave, Morgan Hill, CA, USA) to drive the main mixer. This amplifier effectively suppresses amplitude variations caused by the movement of the flexible RF coaxial cable that connects the power divider to the LO port of the main mixer—a common issue when inserting or removing the fixed attenuator during the measurement process.

To balance the measured and the reference signals detected by the LiA, both the IVD and the PS are adjusted. However, manual balancing becomes increasingly challenging at higher attenuation levels due to larger signal fluctuations caused by noise. To address this issue, automatic interpolation techniques have been introduced, along with the adoption of the programmable IVD as a reference standard. Near the balanced state, the relationship between the IVD setting and the output voltage is assumed to be linear, as illustrated in Figure 2. Let X₁ and X₂ denote the IVD setting values correspond to the negative (V₋) and positive (V₊) output voltages displayed by the LiA, respectively. The IVD value at the balanced state, X, is then calculated using linear interpolation expression as follows.

$$X=X_2-{\displaystyle \frac{\left(X_2{-X}_1\right)V_{+}}{\left|V_{-}\right|{+V}_{+}}}.$$

(3)

This technique significantly simplifies and accelerates the balancing procedure.

To minimize mismatch errors during variable attenuator measurements, precision slide-screw tuners are used in conjunction with three cascaded isolators (ISOL), achieving total isolation greater than 60 dB (i.e., >20 dB per isolator). For fixed attenuator measurements, broadband performance is ensured through a mismatch correction technique that employs 20 dB pad attenuators at each test port [2]. Coiled semi-rigid cables are used to connect the test ports to the DUT, thereby reducing mechanical stress and improving connector repeatability. To suppress undesired leakage coupling, all critical components are enclosed in metal shield cases (as indicated by the dotted lines surrounding the modules in Figure 1) and additionally treated with copper tape and ferrite wave absorbers. Finally, specialized low-pass filters, with attenuation greater than 60 dB from 1 GHz to 25 GHz, are installed at the IF outputs of the mixers to eliminate back-propagating RF signals.

3. Measurement Results and Uncertainty Evaluation

All experiments and measurements described below were conducted in a calibration laboratory under well-controlled environmental conditions, with a room temperature of (23 ± 1) °C and a relative humidity of (50 ± 20)%.

3.1. Optical Fiber Assembly Evaluation

The use of optical fiber assemblies in RF precision measurement systems is relatively uncommon and considered a specialized application. Optical components can introduce specific phenomena such as chirp, polarization effects, and DC drift that may cause instability in the demodulated RF signal [16]. Therefore, in addition to verifying the expected high-isolation performance of the proposed assembly, it is essential to evaluate other potential factors that could degrade overall system performance.

3.1.1. Purity of the Output IF Signal Waveform

Figure 3 shows oscilloscope traces of a 1 kHz IF signal outputs derived from 18 GHz RF measurement signals. The top trace corresponds to the IF output of the reference mixer driven by the signal delivered from the LO source via the proposed optical assembly. This was obtained with the optical signal power, RF drive power, and DC bias voltage to the EOM set to 0 dBm, 10 dBm, and 5.6 V, respectively. The bottom trace shows the IF output of the main mixer driven by the conventional LO signal, for comparison. In both cases, a clean sinusoidal waveform was observed, with no distinguishable difference between them.

3.1.2. Frequency Response and Isolation

The frequency response and isolation performance of the optical fiber assembly were evaluated using a VNA over the 1 GHz to 25 GHz frequency range. The results are shown in Figure 4. The measured parameter |S₂₁|, representing the forward transmission coefficient, exhibits good flatness across the frequency range. In contrast, the reverse transmission coefficient |S₁₂|, which characterizes the isolation performance of the assembly, is not clearly observed, as only the noise floor of the VNA (120 dB) is visible on the screen. This suggests that the assembly achieves an isolation level exceeding 120 dB.

For comparison, achieving an isolation level greater than 120 dB using conventional methods would require considerable hardware complexity. Commercially available coaxial isolators typically offer about 20 dB of isolation each [17]. Thus, at least six isolators must be cascaded to reach 120 dB isolation level. Furthermore, coaxial isolators are usually limited to specific frequency bands. To cover the full frequency range from 1 GHz to 25 GHz, isolators would be required for at least five bands; 1–2 GHz, 2–4 GHz, 4–8 GHz, 8–18 GHz, and 18–26.5 GHz [17]. As each band would require six isolators, a total of approximately 30 isolators would be needed. Such an implementation would result in a highly bulky, complex, and impractical system.

3.1.3. Amplitude and Phase Fluctuations

The amplitude variations of the demodulated RF signal, observed using an RF power meter, were approximately 0.02 dB. While this variation is relatively large and could be a potential source of uncertainty, it does not directly affect the measured signal in the system. This is because the modulated RF signal is used solely to drive the local port of the reference mixer, which operates in saturation and is therefore largely immune to input signal amplitude fluctuations. Figure 5 presents the in-phase component of the IF output signal of the LiA when the system is in a null-balanced state, demonstrating the system amplitude stability. The RF signal level at the main mixer input was −30 dBm and measurement frequency was 25 GHz. The balanced signal was monitored over a 5 min period, during which the IF signal level was intentionally varied by approximately 0.01 dB every minute using the IVD to estimate the extent of signal fluctuation and drift. Under these conditions, the observed signal fluctuations are less than 9.0 × 10⁻⁴ dB and can be considered measurement uncertainties due to amplitude variations. The signal drift over a 5 min period is approximately less than 7.0 × 10⁻⁴ dB, which is considered a negligible source of uncertainty, as a single measurement point is acquired within only a few seconds. Similarly, the phase fluctuation was assessed by monitoring the quadrature component of the IF output signal at the LiA while the system was in a null-balanced condition. Over a 1 min observation period, the phase variation was within ±0.5 degrees. The effect of these phase fluctuations ΔA_ϕ on the signal stability of the system can be estimated using the following equation, expressed in decibels.

$$\begin{gathered} {∆A}_{\varphi }=-20{\mathit{log}}_{10}\left\{{\displaystyle \frac{V\pm V(1-cos\varnothing )}{V}}\right\} \\ =-20{\mathit{log}}_{10}\left\{1\pm \left(1-cos\varphi \right)\right\}, \end{gathered}$$

(4)

where V denotes the voltage amplitude of the measurement signal. Then, ΔA_ϕ can be estimated as ±(3.3 × 10⁻⁴) dB.

3.2. System Performance and Uncertainty Source Summary

The measurable attenuation range, or dynamic range, of the system was primarily determined by its linearity characteristics. To evaluate these characteristics, experiments were conducted to assess the linearity of the system, including the main mixer and preamplifier, leakage as well as the influence of noise. Figure 6 presents measurement results obtained at 25 GHz by measuring 20 dB attenuation increments of a variable step attenuator as the DUT. The input power level to the main mixer was varied in 10 dB steps using the adjustable power leveler, as shown in Figure 1. To minimize mismatch effects, 40 dB isolators were installed at both ports of the DUT. The LiA time constant of 300 ms was used, and each measurement point was averaged over 10 repetitions to enhance measurement stability. The horizontal axis represents the input power level to the main mixer, and the vertical axis shows the deviation from the nominal 20 dB attenuation step. All measured attenuation values were normalized with respect to the measurement at −30 dBm. The system demonstrated good linearity, with deviations of approximately 0.001 dB maintained at main mixer input signal levels as low as −100 dBm.

The major sources of uncertainty in this system are listed and described below. The standard, combined, and expanded uncertainties were evaluated using a conventional analytical method based on the Guide to the Expression of Uncertainty in Measurement (ISO-GUM) framework, assuming that Type A and Type B components are uncorrelated [18].

(1)

Calibration of the IVD: The programmable IVD, Model 6430A, is used as the IF reference standard within the system. However, for calibration purposes, only manually operated IVDs are accepted as valid transfer standards by major national metrology institutes. To comply with this requirement, a manual-type IVD, Model 6415A, is prepared and pre-calibrated at 1 kHz with traceability to the Japan national standard for low-frequency voltage ratio. The 6430A is then calibrated by inserting the calibrated 6415A between the IF amplifier and the 6430A (see Figure 1).

Table 1. Calibration results of the 6430A programmable IVD, summarizing the measured voltage ratios (in dB) and associated uncertainties used as attenuation reference standards at a 1 kHz IF.

		Model 6415A			Model 6430A		En Value
Dial Setting		Nominal	Cal. Value	UT (k = 2)	Meas. Result	Us (k = 2)
Initial	Final	[dB]	[dB]	[dB]	[dB]	[dB]	[dB]
1	0.1	20	20.0000	1.7 × 10−5	19.9996	4.8 × 10−4	0.78
1	0.01	40	40.0000	1.7 × 10−4	39.9995	5.6 × 10−4	0.93
1	0.001	60	60.0001	1.7 × 10−3	59.9994	1.8 × 10−3	0.27

presents the calibration results for nominal attenuation values of 20 dB, 40 dB, and 60 dB, along with the expanded uncertainties U_S (coverage factor k = 2), where the standard uncertainties and their components are estimated based on the contributions listed in

Table 2. Sources of uncertainty in the calibration of the 6430A programmable IVD, outlining the contributing factors to the total uncertainty at the 1 kHz IF.

Model 6430A
Nominal	u(X11)	u(X12)	u(X11)	u(X1)
[dB]	[dB]	[dB]	[dB]	[dB]
20	8.7 × 10−6	2.0 × 10−4	1.3 × 10−4	2.4 × 10−4
40	8.7 × 10−5	2.0 × 10−4	1.8 × 10−4	2.8 × 10−4
60	8.7 × 10−4	2.0 × 10−4	1.9 × 10−4	9.1 × 10−4

u(X₁₁): Standard uncertainty of the 6415A calibration, obtained from its calibration certificate. u(X₁₂): Uncertainty due to the loading effect between the 6415A and the 6430A. In this calibration setup, the two IVDs (6415A and 6430A) are connected in cascade. The finite input impedance of the 6430A imposes a load on the 6415A, introducing a measurable loading effect. This effect is evaluated using an equivalent circuit model [5], and the associated standard uncertainty is estimated to be 2.0 × 10⁻⁴ dB. u(X₁₃): Standard deviation of the mean, obtained from 10 repeated measurements. u(X₁): Combined standard uncertainty, calculated as the root-sum-square (RSS) of the individual components u(X₁₁), u(X₁₂), and u(X₁₃). This value is used as the standard uncertainty u(X₁) associated with the 6430A programable IVD calibration.

. To verify the validity of these calibrations, the results and their associated uncertainties were compared with the calibration values and expanded uncertainties of the 6415A (U_T) as specified in its calibration certificate, using the En-value method [19]. All calculated En values were found to be less than 1.0, indicating good agreement with the reference values from the 6415A calibration.

(2)

Nonlinearity: From Figure 6, the nonlinearity limits are determined to be 6.0 × 10⁻⁴ dB for measurements up to 40 dB (−30 to −70 dBm), and 1.5 × 10⁻³ dB for measurements up to 60 dB (−20 to −80 dBm). For measurement ranges exceeding 60 dB, a double-step technique is employed, allowing mixer input levels of 0 dBm or higher to be applied without inducing saturation effects. As a result, the nonlinearity limits are 1.5 × 10⁻³ dB for measurements up to 80 dB (0 to −80 dBm), and 3.5 × 10⁻³ dB for measurements up to 100 dB (0 to −100 dBm). The corresponding standard uncertainties u(X₂) are calculated by assuming a rectangular probability distribution using a divisor of $\sqrt{3}$.

(3)

Amplitude Fluctuation of the Optical Fiber Assembly: As described in Section 3.1.3, the observed signal fluctuations are less than 9.0 × 10⁻⁴ dB. Assuming a rectangular distribution, the corresponding standard uncertainty u(X₃) is calculated to be 2.6 × 10⁻⁴ dB.

(4)

Phase Fluctuation of the Optical Fiber Assembly: As described in Section 3.1.3, the effect of phase fluctuations on signal stability is estimated to be 3.3 × 10⁻⁴ dB. Assuming a U-shaped distribution using a divisor of $\sqrt{2}$, the resulting standard uncertainty u(X₄) is estimated as 2.3 × 10⁻⁴ dB.

(5)

Leakage: Uncertainty due to leakage arises from signals that bypass the DUT or the reference standard, as well as from signals traveling along unintended paths. Both RF/microwave and IF leakage effects exist in the system; however, the IF leakage is rendered negligibly small through the use of shielded cables, shielded cases, and single-point grounding, which effectively suppress earth loop currents. The dominant contribution comes from RF/microwave leakage, which can be further categorized into internal and external leakage components.

Internal Leakage: The internal leakage shown by Figure 1 depends on the isolation of main and reference mixers and the directivity of the coupler besides the isolation of the optical assembly. The mixers isolation and the directivity of the coupler are assumed to be greater than 20 dB. Then, the system isolation is estimated to be higher than 180 dB, i.e., 20 dB (coupler) + 20 dB (reference mixer) + 120 dB (optical fiber assembly) + 20 dB (main mixer), which means that the internal leakage is 80 dB lower than the measurement signal for an attenuation measurement of 100 dB. Then, by (2), the uncertainty limits ΔA_L due to the internal leakage can be estimated to be ±8.7 × 10⁻⁴ dB.

External leakage: The external leakage path refers to a route through which signals are emitted from the system, propagate externally, re-enter the system, bypassing the DUT, and reach the main mixer. Unlike internal leakage, the attenuation of external leakage paths is difficult to estimate, as it depends on uncontrolled propagation conditions. The leakage effects in the RF circuit are influenced by the relative phase between the measurement signal and the leakage signal at the input of the main mixer. Therefore, these effects can be detected by comparing measurement values under different phase conditions. These conditions were introduced by inserting quarter- and half-wavelength extensions between the DUT and the main mixer. At a signal level of −100 dBm, the differences between measured values were not clearly distinguishable due to a measurement dispersion of approximately 4.0 × 10⁻³ dB. However, a nonlinearity of −1.5 × 10⁻³ dB was observed at this level (Figure 6). Based on these results, the upper bound of the leakage effect at −100 dBm is estimated to be ±0.0015 dB.

The corresponding standard uncertainties, u(X₅₋₁), u(X₅₋₂), for both internal and external leakage are calculated by assuming a U-shaped probability distribution, using a divisor of $\sqrt{2}$. According to expression (3), for every 10 dB increase in the signal level, the standard uncertainties can be estimated by dividing the previous value by $\sqrt{10}$.

(6)

Gauge block attenuator: The single-step (normal) measurement technique is used for attenuation measurements up to 60 dB. For measurements exceeding 60 dB, a double-step technique is employed, in which a gauge block attenuator (e.g., 40-dB) is switched into and out of the circuit. This approach helps to maintain low noise influence during high-attenuation measurements. The standard uncertainty, u(X₆), associated with the gauge block was determined through measurements performed using the single-step technique within this system.

(7)

Mismatch: For variable attenuator measurements, the mismatch uncertainty is minimized by tuning the reflection coefficients at both test ports to be less than 0.01. The residual mismatch uncertainty, u(X₇), is then calculated using expression (38) provided in [20]. For fixed attenuator measurements, the mismatch correction is applied. Subsequently, the mismatch uncertainty is calculated using expression (5) provided in [2].

Table 3. Summary of estimated uncertainties in attenuation measurements of a step attenuator at 18 GHz.

Source of Uncertainty		Category	Probability Distribution	Nominal Attenuation [dB]
				20	40	60	80	100
				Standard Uncertainty [dB]
				u(Xi)	u(Xi)	u(Xi)	u(Xi)	u(Xi)
1	Calibration of the IVD	B	Normal	2.4 × 10−4	2.8 × 10−4	9.1 × 10−4	2.8 × 10−4	9.1 × 10−4
2	Nonlinearity	B	Rectagular	2.3 × 10−4	3.5 × 10−4	8.7 × 10−4	8.7 × 10−4	2.0 × 10−3
3	Amplitude Fluctuation of the Optical Fiber Assembly	B	Rectagular	2.9 × 10−4	2.9 × 10−4	2.9 × 10−4	2.9 × 10−4	2.9 × 10−4
4	Phase Fluctuation of the Optical Fiber Assembly	B	U	2.3 × 10−4	2.3 × 10−4	2.3 × 10−4	2.3 × 10−4	2.3 × 10−4
5	Leakage
5-1	Internal Leakage	B	U	3.8 × 10−7	2.4 × 10−6	1.5 × 10−5	9.7 × 10−5	6.2 × 10−4
5-2	External Leakage:	B	U	6.6 × 10−7	4.2 × 10−6	2.7 × 10−5	1.7 × 10−4	1.1 × 10−3
6	Gauge Block Attenuator	B	Normal				2.4 × 10−3	2.4 × 10−3
7	Mismatch	B	U	1.9 × 10−3	2.3 × 10−3	3.7 × 10−3	3.7 × 10−3	3.7 × 10−3
8	SDOM	A	Normal	2.0 × 10−4	3.0 × 10−4	6.0 × 10−4	7.0 × 10−4	7.0 × 10−4
(u) Combined Standard Uncertainty				2.0 × 10−3	2.4 × 10−3	4.0 × 10−3	4.6 × 10−3	5.1 × 10−3
(U) Expanded Uncertainty		k = 2		4.0 × 10−3	4.8 × 10−3	8.0 × 10−3	9.2 × 10−3	1.0 × 10−2

summarizes the estimated uncertainties associated with attenuation measurements of a step attenuator (8496H) at 18 GHz. The standard deviation of the mean (SDOM) was determined from 10 repeated measurements. Uncertainty sources related to the fiber-optic assembly are indicated in italics. The expanded uncertainty is reported as the combined standard uncertainty multiplied by a coverage factor (k = 2), corresponding to a coverage probability of approximately 95% for a normal distribution [18].

The optical fiber assembly was found to significantly reduce uncertainty due to internal leakage. Although it introduces inherent uncertainty factors related to phase and amplitude stability, these effects were minimal and did not represent major contributions to the combined standard uncertainty.

As a form of validation, this system was included in the Consultative Committee for Electricity and Magnetism (CCEM) Key Comparison CCEM.RF-K26 at a frequency of 18 GHz, with a measurement range extending up to 90 dB. The results from our institution, identified by the initials “NMIJ,” demonstrated very satisfactory performance in this comparison [21].

4. Conclusions

An accurate attenuation measurement system covering a wide attenuation range across microwave frequencies from 1 GHz to 25 GHz was developed, based on the dual-channel IF substitution method using an IVD as a reference standard. The system is capable of evaluating both variable and fixed attenuator DUTs, supporting a measurement range exceeding 100 dB. To facilitate automation and precision, a programmable IVD is integrated, along with an accurate null-detection technique based on interpolation. A newly designed optical fiber assembly, incorporating an EOM, enables broadband optical modulation and plays a critical role in suppressing internal leakage effects. This contributes to accurate high-attenuation measurements without introducing significant side effects, while also offering broad frequency coverage and system design flexibility. For high-attenuation conditions, a double-step measurement technique was employed to maintain measurement reliability. A comprehensive performance evaluation was conducted, highlighting the programmable IVD calibration technique, which achieved an accuracy better than 0.001 dB across all attenuation levels, and the effectiveness of the optical assembly, which provided isolation levels exceeding 120 dB across the entire frequency range. This high isolation significantly contributes to suppressing internal leakage effects to below 0.001 dB at 100 dB attenuation. The system achieves expanded uncertainties (k = 2) of 0.004 dB at 20 dB, 0.008 dB at 60 dB, and 0.010 dB at 100 dB attenuation. A technical manual for operating the system as both a primary standard and a quality system under ISO/IEC 17025 was been prepared [22]. The system was successfully utilized in the CCEM Key Comparison CCEM.RF-K26, where it demonstrated satisfactory performance. Moreover, the system can be scaled to higher frequencies, such as the Ka-band or millimeter-wave region, by replacing commercially available RF and optical components in accordance with the target frequency band.